Study Of Vehicle Positive Feature Recognition Method

Vehicle feature recognition has been the intelligent traffic management and the main research directions, and has wide application and development space, at the same time these recognition technology involves many disciplines of theoretical knowledge, have important theoretical significance.In this paper, the research object is in the camera is fixed under the collected traffic image data, the image analysis and processing, real time extraction of the frontal vehicle characteristics, for recognition. Recognition is the main purpose of traffic statistics and vehicle-logo recognition, according to different experimental data for different application.According to the experimental data and objective analysis, for the light is not full scene video images this paper proposed a luminance difference and extraction of connected region combination method for the detection and localization of motor vehicle headlights, the method can accurately detect and locate on the road still and motion of the vehicle, but also can the effective removal of rear lamp irradiation of light beam impact, according to the detection and location of image results in car traffic statistics; for light full scene images using Otsu (Otsu method) and the second level gradient projection method of combining the image of vehicle logo, the detection and recognition, can accurately detect light with position, orientation of vehicle-logo location. These research content for the later moving vehicle detection, localization and vehicle-logo recognition method is provided for the research value.Road red light appears when the queue is special, with around two car headlights position basically unchanged method to deal with image, which were detected by using this method, positioning, very high reliability; in vehicle logo detection, the Otsu method is improved to achieve the image threshold segmentation, extraction of lamp strip the location, on lamp belt in the logo for further testing. This two approach involves threshold segmentation can effectively remove the interference of image objects, and the traditional frame difference and background difference effect is not so good. Using the threshold segmentation to extract the target location, for more accurate detection and localization. From the experimental results it can be stated in this paper the method in the corresponding scene, accurate detection, can achieve real-time requirements.According to the different requirements of vehicle feature recognition, in the VC platform called OpenCV standard library function C language applications to identify the requirements for a programming. And this algorithm is applied to the actual traffic scene acquisition vehicle image in the corresponding testing:statistical vehicle flow, identification mark and related transportation vehicle characteristics, and according to the test contents and results were given different under traffic scene experimental statistical data and performance evaluation.